Representing Isabelle in LF

LF has been designed and successfully used as a meta-logical framework to represent and reason about object logics. Here we design a representation of the Isabelle logical framework in LF using the recently introduced module system for LF. The major novelty of our approach is that we can naturally represent the advanced Isabelle features of type classes and locales. Our representation of type classes relies on a feature so far lacking in the LF module system: morphism variables and abstraction over them. While conservative over the present system in terms of expressivity, this feature is needed for a representation of type classes that preserves the modular structure. Therefore, we also design the necessary extension of the LF module system.

💡 Research Summary

The paper presents a novel encoding of the Isabelle logical framework within the LF (Logical Framework) using the recently introduced LF module system, and extends that system with a new feature—abstraction over morphism variables—to faithfully represent Isabelle’s advanced modular constructs, namely type classes and locales. The authors begin by observing that both Isabelle and LF are based on the Curry‑Howard correspondence, allowing proofs to be treated as terms, yet Isabelle’s module system (theories, locales, type classes, imports, sublocales, interpretations, subclassing, and instantiations) is richer than what the original LF module system can directly capture.

In the background sections the paper briefly reviews Isabelle’s core syntax (a simplified grammar is given) and LF’s basic notions (signatures, type families, constants, dependent function types Π, and the Twelf implementation). The original LF module system supports signatures, inclusion, structures, and views, but it lacks the ability to treat signature morphisms as first‑class entities that can be parameterised or abstracted over. This limitation prevents a direct, structure‑preserving translation of Isabelle’s type‑class hierarchy, where a class may inherit from other classes and instances are obtained by applying a morphism that substitutes parameters.

To overcome this, Section 3 introduces “morphism variables” (also called morphism abstraction). A morphism variable can be bound in a signature, passed as an argument, or returned as a result, much like a term variable. The authors formalise the typing rules for such variables, prove that the extension is conservative (it does not increase the expressive power of LF beyond what can already be encoded, but it enables more natural encodings), and show how to integrate them with existing LF constructs such as structures and anonymous morphisms. This addition effectively makes the LF module system capable of representing parametrised signatures and their instantiations in a way that mirrors Isabelle’s module semantics.

Section 4 then gives the concrete encoding. Every Isabelle theory is mapped to an LF signature; locales become parametrised LF signatures; type classes are treated as a special subclass of locales with a single type variable constrained by a class identifier. Relations between modules are encoded as LF signature morphisms: imports become inclusion morphisms, sublocales and interpretations become morphisms that map the source locale’s symbols to the target’s symbols (potentially with parameter substitution), subclassing is represented by morphism composition, and instantiation of a type class is modelled by applying a morphism variable that supplies concrete definitions for the class operations. The authors illustrate the encoding with a running example involving ordering and semilattice classes, showing how the LF signatures and morphisms faithfully capture inheritance, parameter passing, and proof obligations.

A key insight is that the LF representation preserves the modular structure of Isabelle: sharing, reuse, and hierarchical relationships are reflected directly in the LF signature graph and its morphism network. Because LF has a well‑understood formal semantics, this translation provides a meta‑theoretic foundation for reasoning about Isabelle modules, and it opens the possibility of using LF as an intermediate language for exporting Isabelle libraries to other proof assistants. The paper acknowledges that an actual implementation of the translation pipeline is future work, but the formal design already demonstrates feasibility.

In conclusion, the authors make two substantial contributions: (1) they extend the LF module system with morphism abstraction, a modest yet powerful addition that enables first‑class treatment of signature morphisms; (2) they show how this extended system can encode Isabelle’s sophisticated module mechanisms, preserving their modularity and enabling LF‑based reasoning about Isabelle developments. This work bridges a gap between two influential logical frameworks and suggests a path toward greater interoperability among theorem proving environments.